[DX12] Panorama skybox crashes editor

[Dadaskis]

Tested versions

• Reproducible in 4.3.stable.

System information

Godot v4.3.stable - Windows 10.0.22631 - d3d12 (Forward+) - dedicated NVIDIA GeForce GT 1030 (NVIDIA; 32.0.15.6081) - Intel(R) Core(TM) i5-7400 CPU @ 3.00GHz (4 Threads)

Issue description

So... I have a PanoramaSkyMaterial

And i have a fancy lossless image i want to use as a skybox...

And when i drag this image over a Panorama variable to set it as a skybox texture - it crashes the editor.

It happens only when i use DX12.

Why am i using DX12? Because i experience a severe lag on Vulkan, so, yeah, i need that

Steps to reproduce

• Download MRP
• Launch the project, go to test.tscn
• Locate WorldEnvironment
• Repeat steps mentioned in the description
• Crash

Minimal reproduction project (MRP)

skyboxcrash.zip

[AtlaStar]

was able to reproduce

Attempted to open project in 4.3 branch and met with this access violation error caused by trying to dereference a null pointer

EDIT: looks like this is just a matter of me not configuring things correctly, as I didn't build godot with the appropriate flag to enable dx12 as is required in the display server here. Gonna rebuild and test.

Perhaps this behavior warrants a report as one would expect loading a project to not crash, but rather fallback to a different rendering server if the one set in the project isn't available?

[AtlaStar]

So, after building things properly, unable to reproduce with the following; Godot v4.3.1.rc (6699ae7) - Windows 10.0.19045 - d3d12 (Forward+) - dedicated AMD Radeon RX 6650 XT (Advanced Micro Devices, Inc.; 32.0.11027.1003) - AMD Ryzen 7 3800X 8-Core Processor (16 Threads)

Going to compare this build to stable, then perform a bisection if necessary.

[AtlaStar]

Alright, reproduced for real this time

Gonna bisect because it seems like the issue may have to do with mipmap generation if I understand correctly at least, as it appears that the generation can produce dimensions of 0 as shown here.

I may be misunderstanding the core issue though.

EDIT: to clarify my thoughts though, the code where the error is occurring hasn't changed since implemented according to the blame, and the fact it appears to be a regression suggests that it has to do with either the resource itself, or how it tries to load/bind the resource to the node.

[AtlaStar]

So, bisect shows that this is a duplicate of #96352 and is in fact a mipmap issue. Fix is already in 4.4, and was in fact cherry picked for 4.3.1.rc as well.

[clayjohn]

Thanks for bisecting!

Closing as already fixed!